Google Releases Offline AI Meeting Assistant for Mac

Google has introduced AI Edge Foresight, a new AI-powered meeting note-taking application designed to work completely offline on Mac computers. The experimental app uses Google’s EmbeddingGemma 2 model, a 740-million-parameter multimodal model designed to process information locally. This allows Foresight to capture meeting conversations, generate notes and work with other information without sending the user’s meeting data to the cloud.

Foresight is designed around a split-screen workspace where users can write shorthand notes while the AI generates more detailed notes from the meeting transcript. The application can also provide a transcript of the conversation and allow users to ask questions about what was discussed. Google says the system can work with both online meetings and conversations taking place in person, making it useful even when there is no internet connection.

The application can also create a personal knowledge base from local files. Users can provide documents such as PDFs, Google Docs, Microsoft Office files, plain-text documents, Markdown files and web bookmarks. Foresight can then use information from those materials together with meeting notes to answer questions and provide relevant information during or after a meeting.

Privacy is one of the main advantages of the system. Because the AI processing is performed locally, meeting audio, transcripts, notes and files can remain on the user’s computer rather than being uploaded to a remote AI service. Google says the application is optimized specifically for Apple Silicon Macs, allowing the local model to run without depending on cloud computing infrastructure.

The launch also demonstrates Google’s broader push toward smaller AI models capable of running directly on consumer hardware. EmbeddingGemma 2 is an open-weight model released under the Apache 2.0 license and is designed for local multimodal search across text, images, video and audio. With Foresight, Google is showing how increasingly capable AI can be moved from cloud servers onto personal computers, potentially improving privacy, reducing latency and allowing AI tools to function even without an internet connection.